Not working

[Juphex]

I did something like this:

#convert to libsvm format https://github.com/zygmuntz/phraug/blob/master/csv2libsvm.py
#python csv2libsvm.py spambase_renamed_class.csv libsvm.data -1 False

data = open("spambase/libsvm.data", "r")
data2 = open("spambase/libsvm2.data", "a")

i = 0
for line in data:
    temp = line[-3:]
    white_space_index = line[2:].find(" ")
    
    if i == 0:
        white_space_index = 0
        i += 1
    
    if "-1" in temp:
        line = "-1 " + line[white_space_index + 2:-6]
    else:
        line = "1 " + line[white_space_index + 2:-6]

        #drop last index (class) and first because it is the index nr.
    data2.write(line + "\n")
data.close()
data2.close()

[Juphex]

When you know the highest occuring index you can do something like this:

data = open("iris.scale", "r")
def read_libsvm(data, col_length):
    new_data = np.zeros((1,col_length))
    i = 0
    for line in data:
        label = line[:1]
        #split splits on whitespace by default
        x = line[1:].split()
        if i > 0:
            new_data = np.vstack((new_data, np.zeros((1,col_length))))
        for j in range(len(x)):
            temp_string = x[j].split(":")
            new_data[i, int(temp_string[0]) - 1] = np.float32(temp_string[1])
        new_data[i, col_length - 1] = int(label)
        i += 1
    return new_data